Development and progression of prostate cancer is associated with the accumulation of complex genetic and epigenetic aberrations.1 Recently, we identified common allelic imbalance at 6q14-6q22 in prostate cancer by a genome-wide loss of heterozygosity (LOH) analysis using high-density single nucleotide polymorphism (SNP) microarrays.2 Other groups have earlier reported similar regions of deletion/LOH at 6q in prostate cancer (Ref.3 and references therein); however, a possible target tumor suppressor gene (TSG) remains to be identified from this region. Interestingly, several cancer types, including breast,4 gastric5 and cervical cancer6 as well as a variety of hematopoietic malignancies7, 8 have common regions of deletion at 6q that coincide with that identified in prostate cancer, suggesting that a putative target TSG at 6q may be shared by at least some of these malignancies.
Retroviral tagging has proved to be a powerful cancer gene discovery tool, with many genes identified as common integration sites in retrovirus-induced murine leukemia/lymphoma models, playing critical roles also in human cancers, including solid tumors.9, 10 On the basis of this and substantiated by the overlapping deletions at 6q found in both prostate cancer and various types of human hematopoietic malignancies, we envisaged that retroviral tags from murine leukemia/lymphoma models might help the search for a candidate TSG at 6q. Hence, in this study, to identify a new possible TSG in prostate cancer, we hypothesized that a gene in the main LOH region at 6q14-6q22 that is both (i) transcriptionally downregulated in prostate cancer compared to nonmalignant prostate tissues and (ii) disrupted by retroviral insertion in murine hematopoietic cancer models, would provide a good candidate for further analysis. Accordingly, the SRC family tyrosine kinase gene FYN at 6q21 was selected for further investigation.
FYN contains an N-terminal SH4 myristoylation and palmitoylation domain, which targets the protein to the inner plasma membrane. It is followed by a unique sequence, distinguishing FYN from other SRC family members, 2 protein–protein interaction domains (SH3 and SH2), a kinase domain (SH1) and a C-terminal negative regulatory domain.11 The 2 known major splice forms, FYN(B) (aka P59FynB) using exon 7A and FYN(T) (aka P59FynT) using exon 7B,11 differ in a linker region between SH2 and SH1 that extends into the kinase domain, causing moderate regulatory differences between these isoforms.12 FYN(B) is highly expressed in brain, and hematopoietic cells express mainly FYN(T), while most other tissues express both isoforms.11 A variant mRNA lacking exon 7 (FYN(Δ7)) exists in blood cells, but the corresponding endogenous protein has not been documented.13
The biological functions reported for FYN are diverse, including both stimulatory and inhibitory effects on cellular differentiation, proliferation and survival.11 FYN signaling is required for activation of T-cells through the T-cell receptor (TCR) and for FAS- and TCR-mediated apoptosis in these cells.14, 15 Furthermore, a tumor suppressor role has been suggested for FYN in neuroblastomas, possibly through induction of differentiation and cell cycle arrest.16 Recently, hypermethylation of the FYN gene promoter was detected by methylation-specific PCR in breast17 and gastric carcinoma cell lines,18 indicating that FYN is regulated also at the epigenetic level. In the present work, which is the first to describe FYN as a candidate TSG in prostate cancer, we have found that FYN expression is downregulated in several prostate adenocarcinoma cell lines and in primary prostate cancer samples compared with nonmalignant prostate cells and tissues. Furthermore, we have identified 2 possible mechanisms responsible for this, chromosomal deletion and promoter hypermethylation.
Material and methods
Benign prostate hyperplasia (BPH) and prostate adenocarcinoma specimens were obtained as transurethral resections or as needle biopsies from patients undergoing radical prostatectomy. Tissue samples were fresh frozen or embedded in Tissue Tek (Bayer Corporation, Pittsburgh, PA) and stored at −80°C. Formalin-fixed and paraffin-embedded (FFPE) tissue samples were also available. Clinicopathological information is provided in supplementary Table S1. Informed content was obtained from all patients in this study, which was approved by the Scientific Ethics Committee of Aarhus County, Denmark.
Laser-microdissection and SNP microarray analysis
Laser-microdissection and SNP microarray analysis have been previously described.2 In short, adenocarcinoma cells were microdissected from 5-μm hematoxylin-stained sections of fresh frozen or Tissue Tek-embedded prostate cancer specimens using the PALM Microbeam System (PALM Microlaser Technologies AG, Bernried, Germany). Genomic DNA was isolated from 40 microdissected prostate cancer specimens and matching blood samples (normal DNA) using the PUREGENE DNA Purification Kit (Gentra Systems, Minneapolis, MN), and analyzed on SNP microarrays covering 58,960 SNPs (GeneChip Human Mapping 50K Array Xba 240; Affymetrix, Santa Clara, CA). Microarray experiments were carried out according to standard protocols provided by Affymetrix. Genotype calls were generated in GeneChip® DNA Analysis Software 3.0.2 (Affymetrix) and visualized using dChip software (http://www.dchip.org).
Cell culture and 5-aza-2′-deoxycytidine treatment
All cell lines were grown in RPMI 1640 with L-glutamine (Gibco, Invitrogen Corporation, Carlsbad, CA) supplemented with 10% fetal bovine serum, 100 U/ml penicillin and 100 μg/ml streptomycin. LNCaP, PC-3, DU145 and 22rv1 prostate adenocarcinoma cells were purchased from ATCC and BPH-1 cells from DSMZ (Braunschweig, Germany). Dr. Kenneth Pienta, University of Michigan, kindly provided VCaP and DuCaP prostate adenocarcinoma cells. PSK-1 prostate small cell carcinoma and 1013L bladder transitional carcinoma cells were kind gifts from Dr. Adrie van Boekhoven, University of Colorado. PNT1A immortalized prostate epithelial cells are described in Degeorges et al.19 All cell lines were treated for 48 hr with 1 μM 5-aza-dC (Sigma-Aldrich Corporation, St. Louis, MO) and allowed 5-days recovery in complete medium prior to RNA, DNA and protein extraction. Untreated cells were grown in parallel. All experiments were performed in duplicate.
Protein extraction and Western blotting
Protein extracts from cultured cells or 5-μm sections of fresh-frozen prostate tissue samples were prepared in IEF lysis buffer (9.8 M urea, 2% NP40, 2% carrier ampholytes pH 7–9, and 100 mM DTT). Specific sections containing areas of adenocarcinoma, BPH or nonmalignant adjacent glands only were carefully selected. Protein concentrations were determined by the Bradford method (Dye Reagent Concentrate; Bio-Rad, Hercules, CA) with BSA as standard. For Western analysis, protein lysates were run on 4–12% NUPAGE polyacrylamid gels (Invitrogen Corporation), blotted to PVDF membranes (Immobilon-P Transfer Membrane; Millipore, Billerica, MA) and blocked with 5% skimmed milk and 0.1% Tween-20 in PBS buffer. The following primary antibodies were used: mouse monoclonal anti-FYN (clone 25; BD Transduction Laboratories, San Jose, CA), mouse polyclonal anti-FYN (A01; Abnova Corporation, Taipei, Taiwan), mouse monoclonal anti-FYN (ab3116; Abcam, Cambridge, UK) and mouse monoclonal anti-ACTIN-β (Sigma-Aldrich). HRP-conjugated goat anti-mouse IgG (Dakocytomation, Glostrup, Denmark) was used as secondary antibody, and bands were visualized with the ECL plus WB Detection System (Amersham Biosciences, Freiburg, Germany). Lysate from PNT1A cells transfected with an expression vector for full-length human FYN(B) (pCMV5-huFYN; kind gift from Dr. Marilyn Resh, Memorial Sloan-Kettering Cancer Center, New York) was used as a positive control.
Immunohistochemical analyses of FFPE prostate tissue sections and prostate adenocarcinoma tissue microarrays (BioCat, Heidelberg, Germany) were performed as described in Ref.20, except that citrate buffer (10 nM, pH 6.0) was used for FYN epitope demasking. Antibodies against FYN (clone 25, BD Transduction Laboratories) and high-molecular-weight cytokeratin (HMW-ck; Dakocytomation) were diluted 1:10 and 1:50 in TBS (50 mM Tris; 0.9% NaCl, pH 7.6) with 1% BSA. The anti-mouse EnVision+ System with HRP-labeled polymer (Dakocytomation) and DAB solution (Kem-En-Tec, Copenhagen, Denmark) was used for secondary staining.
RNA preparation, cDNA synthesis and (real-time) RT-PCR
The RNeasy Micro Kit (Qiagen, Valencia, CA) was used to isolate total RNA from cultured cells and carefully selected Tissue Tek-embedded prostate tissue specimens (25–30 twenty-micrometer sections). First-strand cDNA synthesis was performed with SuperScript II Reverse Transcriptase (Invitrogen Corporation) using an oligo(dT)24 primer. Reverse transcriptase PCR (RT-PCR) was performed with primers in exon 6 (FYN-exon6-F) and exon 8 (FYN-exon8-R) of the FYN gene, and amplicons were analyzed on agarose gels before and after XhoI digestion. Real-time RT-PCR reactions were run in triplicate on the ABI PRISM 7000 Sequence Detection System (Applied Biosystems, Foster City, CA), with quantitation based on standard curves. Total FYN expression was determined with a TaqMan Gene Expression Assay detecting all FYN isoforms (Hs00176628_m1, Applied Biosystems). In addition, primers specific for FYN(B) (Fyn-exon6-F+FYN-exon7A-R) and FYN(T) (FYN-exon7B-F+FYN-exon7B-R1), respectively, were used for real-time RT-PCR with SYBR GREEN PCR Master Mix (Applied Biosystems). CFL1 (COFILIN-1) was used for normalization. Primer sequences are given in supplementary Table S2.
Genomic DNA from prostate cell lines and carefully selected 20-μm sections of BPH, prostate adenocarcinoma and adjacent nonmalignant prostate tissue was isolated with the PUREGENE™ DNA Purification Kit (Gentra Systems) and bisulfite-converted using the MethylEasy DNA Bisulfite modification kit (Human Genetic Signatures, Sydney, Australia). DNA from 5-aza-dC treated cells was used as a negative control and CpGenome Universal Methylated DNA (Chemicon, Temecula, CA) as a positive control for methylation. FYN promoter CpG island sequences were amplified from bisulfite-converted DNA by PCR, purified from agarose gels (QIAquick Gel Extraction Kit, Qiagen) and subcloned into the pCR4-TOPO vector (TOPO TA Cloning Kit for Sequencing, Invitrogen). Individual clones were sequenced with M13 primers using the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems) and analyzed on an automated ABI PRISM 3100 Genetic Analyzer (Applied Biosystems). Furthermore, bulk PCR products were sequenced directly in order to screen for the presence of methylated CpGs prior to analysis of individual clones. Primer sequences are provided in supplementary Table S2.
Candidate gene selection
To identify a new possible prostate TSG candidate, we used the selection criteria described earlier. Of 201 annotated genes (NCBI, Human Genome Build 36.1) in the common LOH region at 6q14-6q22 found in prostate cancer (Fig. 1a), the murine homologs of 12 were disrupted by retroviral integration(s) (i.e., suspected loss-of-function mutations) in mouse tumors, based on data from the Retrovirus Tagged Cancer Gene Database (http://rtcgd.ncifcrf.gov; Ref.21) (not shown). Viral integrations outside genes were excluded, because they most likely represent proto-oncogene activating mutations. Three of the 12 genes were hit multiple times (Bach2, Prdm1, Sclm4), whereas the remaining 9 were hit once in the database. As judged from microarray expression data in the Oncomine Cancer Profiling Database (http://www.oncomine.org; Ref.22), 2 of the 12 corresponding human genes (FYN and SNX14) were downregulated at the RNA level in prostate cancer compared with normal prostate tissue and/or BPH, with FYN being more significant (Fig. 1b). It was therefore chosen as our prime candidate for further analysis.
Low FYN expression levels in prostate adenocarcinoma cell lines
To investigate FYN expression patterns in prostate cells, total RNA was isolated from 12 prostate tissue samples (6 adenocarcinomas, 4 BPHs and 2 adjacent nonmalignant), 6 prostate adenocarcinoma cell lines (LNCaP, PC-3, DU145, 22rv1, VCaP and DuCaP), PNT1A immortalized prostate epithelial cells and BPH-1 cells. PSK-1 prostate small cell carcinoma and 1013L bladder transitional carcinoma cells were also included because of their high FYN expression levels (see later).
A primer set flanking exon 7 was used for RT-PCR analysis to simultaneously detect all 3 FYN splice variants. All investigated cell lines expressed FYN(B) and FYN(T), but FYN(Δ7) was not detected (Fig. 2a). In addition to FYN(B)- and FYN(T)-specific bands, primary prostate cancer, adjacent nonmalignant and BPH samples in most cases produced a weak band of 181 bp specific for FYN(Δ7) (Fig. 2b, and not shown). This band may derive from contaminating blood cells in the crudely dissected tissue samples. Next, FYN expression in the cell lines was quantitated by real-time RT-PCR using a commercial TaqMan gene expression assay that detects all known FYN splice variants, plus 2 primer sets designed specifically for FYN(B), respectively, FYN(T) transcripts. The highest level of total FYN was measured in 1013L cells, followed by PSK-1 and BPH-1 cells (Fig. 2c). Compared to BPH-1 cells, total FYN expression was significantly lower (3–14-fold) in all 6 adenocarcinoma cell lines. Moreover, 3 of these (LNCaP, PC-3 and DU145) showed at least 2-fold lower total FYN expression than PNT1A cells (p < 0.05; t-test) (Fig. 2c), representing an intermediate stage between normal and malignantly transformed epithelial cells.19 Except for the relatively low FYN(T) transcript levels in PNT1A and 1013L cells, FYN(B) and FYN(T) expression followed the same overall patterns as total FYN in all cell lines (Fig. 2c).
FYN protein levels determined by Western blotting analysis of crude cell line extracts paralleled the total FYN mRNA levels measured by real-time RT-PCR. Using a monoclonal antibody (BD Transduction Laboratories) raised against the N-terminal fourth of human FYN, identical in all 3 isoforms, we detected a band of ∼59 kDa; the expected size of full-length FYN(B) and FYN(T) (Fig. 2d). The same band pattern was seen with a polyclonal antibody (Abnova) raised against full-length human FYN(Δ7) (not shown). Although, this does not formally prove that FYN(B) and FYN(T) proteins are both present in the ∼59-kDa band, we note that both transcript variants were found in all cell lines (Fig. 2a). Commercial FYN isoform-specific antibodies were not available.
Reduced FYN protein expression in prostate cancer tissue
The downregulation of FYN mRNA in prostate cancer tissue has been shown in several studies using microarrays23, 24, 25, 26, 27, 28 (Fig. 1b), which do not distinguish between FYN splice variants. Using real-time RT-PCR, we found that total FYN expression was reduced 2-fold in prostate cancer compared with that in BPH tissue samples (Fig. 3a). Although not statistically significant in our sample set, a similar trend was measured for FYN(B) and FYN(T) transcript levels, suggesting that both splice variants are downregulated in prostate cancer (Fig. 3a). Likewise, by Western blotting analysis of protein extracts from crudely dissected prostate tissue specimens, we found ∼2-fold lower levels of FYN protein in prostate cancer tissue than in BPH samples (Figs. 3b and 3c). In addition to a weak band of ∼59 kDa, corresponding to the size of full-length FYN(B)/FYN(T), a more intense band of ∼53 kDa was detected in all tissue samples, along with another weak band of ∼56 kDa in some of the tumors (Fig. 3b). This band pattern was seen with 3 different anti-FYN antibodies, the 2 used earlier for the cell lines, plus another monoclonal antibody that does not cross-react with other SRC family members (Abcam), strongly arguing that these bands are specific (data not shown). Such additional FYN-specific bands have previously been reported by others using Western analysis and were found to represent proteolyzed FYN protein.12, 29 The low FYN(Δ7) transcript levels in BPH and prostate cancer tissues (Fig. 2b) make it less likely that the intense ∼53-kDa band corresponds to FYN(Δ7) protein. Furthermore, no mutations that could possibly explain the smaller sized band(s) were found in the coding sequences of FYN(B) and FYN(T) transcripts amplified from these samples (not shown).
To further investigate FYN protein expression, we used a prostate adenocarcinoma tissue microarray for immunohistochemistry. Normal and adjacent nonmalignant prostate glands contain a basal cell layer, which is absent in prostate cancer. Therefore, staining for the basal cell marker HMW-ck was performed in parallel to distinguish adjacent nonmalignant glands from cancerous glands. Positive staining for FYN was detected exclusively in nonmalignant glands, also positive for cytokeratin, whereas FYN protein was undetectable in nearby prostate adenocarcinoma tissue, negative for cytokeratin (Figs. 4a–4d). This pattern was highly consistent, although in some cases FYN expression was below the detection limit even in nonmalignant adjacent glands. The samples on the tissue microarray were divided into 2 groups based on their cellular composition: 22 cores had high cancer content (>90% of the area contained cancer cells, <10% of the area contained nonmalignant HMW-ck+ glands), while 31 cores had low cancer content (≤90% of the area contained cancer cells, ≥10% of the area contained nonmalignant HMW-ck+ glands). All sections (22 out of 22) with high cancer content were FYN negative, whereas FYN staining (usually heterogeneous) was detected in 21 of the 31 (68%) sections with low cancer content (and high content of nonmalignant HMW-ck+ glands) (Figs. 4a and 4c, and not shown). This difference is highly significant (p < 10−5; Fisher's exact test). Normal prostate epithelium displayed a more homogeneous FYN staining pattern (Fig. 4e), suggesting that downregulation is already initiated in adjacent nonmalignant glands, before loss of the basal cell layer. Despite normal histological appearance, adjacent nonmalignant glands have been shown to differ from truly normal glands at the molecular level.30
Aberrant methylation of a FYN promoter CpG island in cell lines with low FYN expression
Hypermethylation of cytosines in CpG islands is associated with transcriptional silencing of TSGs in cancer cells.31FYN contains a 1.7-kb CpG-rich sequence at its promoter/5′ region (Fig. 5a), which has been found hypermethylated in breast17 and gastric carcinoma cell lines.18 To investigate possible epigenetic regulation of FYN in prostate cancer cells, the CpG island region was analyzed by genomic bisulfite sequencing in the 3 adenocarcinoma cell lines with the lowest endogenous FYN expression levels (LNCaP, DU145, PC3), in 2 adenocarcinoma cell lines with medium level FYN expression (DuCap and VCaP), and in 2 prostate cell lines with high endogenous FYN expression (PSK-1 and BPH-1) (see Fig. 2c). A total of 147 CpGs were interrogated from 4 distinct regions (termed region I to IV) of the CpG island (Fig. 5a).
The CpG island was almost completely free of methylation in PSK-1, BPH-1, DuCaP and VCaP cells (Fig. 5a), indicating that differences in endogenous FYN expression between these cell lines are not caused by promoter methylation. In contrast, low density (7%) methylation was detected in LNCaP cells, in which particularly regions I and IV were affected (15% methylation). In DU145 and PC3 cells, even higher density aberrant methylation was detected in region IV (29 and 37%, respectively), while regions I–III remained largely unmethylated (0–4%), giving a total average methylation level of 10 and 13%, respectively (Fig. 5a). Together, these findings suggest that region IV is a preferred region for aberrant methylation in prostate adenocarcinoma cells.
As little as 6–8% methylation has been shown to cause significant downregulation of TSGs, although higher densities (>20–40%) are required for total silencing.32, 33 Accordingly, the methylation levels detected in DU145 and PC3, and perhaps even LNCaP, could be sufficient for downregulation of FYN expression in these cell lines. To investigate this possibility, FYN expression was measured by real-time RT-PCR following treatment of the cells with the DNA methylation inhibitor 5-aza-dC. FYN expression was significantly induced by 5-aza-dC treatment in DU145 and PC3 (Fig. 5b), suggesting that demethylation of region IV may lead to upregulation of FYN in these cell lines. In contrast, FYN expression was not induced in LNCaP cells, indicating that the low methylation density in this cell line has no influence on expression. FYN expression was also not significantly affected by 5-aza-dC treatment in PSK-1, BPH-1 and DuCaP cells (Fig. 5b), in agreement with the lack of methylation in these cell lines (Fig. 5a). However, in VCaP cells, which are also free of methylation, 5-aza-dC treatment caused significant induction of FYN expression, indicating that FYN upregulation is indirect in these cells. Although such indirect effects may be cell line specific, it cannot be excluded that they play a role also for the induction of FYN expression by 5-aza-dC in DU145 and PC3 cells. We note, however, that DU145 and PC3 had the highest densities of methylation as well as the lowest endogenous FYN expression levels of all cell lines investigated, showing that at least in some cell lines there is an inverse correlation between expression and promoter methylation levels of FYN. Two subregions appeared particularly prone to methylation in both DU145 and PC3 cells (CpG no.11–19 and CpG no.32–37 of region IV).
FYN promoter hypermethylation in prostate cancer tissue
To investigate possible FYN promoter hypermethylation in primary prostate adenocarcinomas, genomic DNA isolated from 3 BPHs, 4 adjacent nonmalignant and 18 prostate cancer tissue samples was analyzed by bisulfite sequencing. Initially, we focused on region IV, which was aberrantly methylated in both DU145 and PC3 cells (Fig. 5a). The 7 nonmalignant tissue samples were all found to be unmethylated, whereas 12 out of 18 tumor samples contained DNA molecules with methylation in region IV of highly variable density (range: 12–95% methylated CpGs) (Fig. 6). Unmethylated clones in the tumors most likely derive from normal cell DNA contamination, the extent of which varies between our samples (compare e.g., PC-57 and PC-41, Fig. 6). In the remaining 6 out of 18 tumor samples (PC-06, 21, 30, 45, 53 and 59), significant aberrant methylation of region IV was not detected (Fig. 6). These results indicate that aberrant methylation of region IV, usually of moderate density, is a frequent event in prostate cancer, as was also seen in prostate adenocarcinoma cell lines (Fig. 5).
Next, to screen for the presence of methylated CpGs in regions II–III, located around the FYN transcription start site, bulk PCR products from the same 25 tissue samples were analyzed by bisulfite sequencing without subcloning. This screening method was highly sensitive for the detection of methylated CpGs in a mixture of DNA molecules and produced very few false positives, as verified by a systematic comparison of bulk bisulfite sequencing results for region IV (data not shown) with sequences of the corresponding individual clones in 25 samples (Fig. 6). Screening of regions II and III detected no methylated CpGs in the 7 nonmalignant samples investigated, whereas 2 of 18 tumor samples had several methylated CpGs (PC-20 and PC-57) (not shown). The remaining 16 tumors were free of methylation in regions II–III (not shown). For further analysis, several individual clones were sequenced from the 2 tumors screened positive (Figs. 7a–7b) and from 2 representative samples screened negative for methylation (Figs. 7c and 7d). Consistent with bulk sequencing results, DNA molecules with dense hypermethylation of region II and III were detected in both PC-20 and PC-57, while all clones were unmethylated in PC-600 and PC-30.
In conclusion, we have found that aberrant methylation of region IV, usually of moderate density, is a common, tumor-specific event in prostate cancer (12 of 18 tumors = 67%), while dense hypermethylation of the entire FYN promoter CpG island (consistent with gene silencing31) occurs less frequently (2 of 18 tumors = 11% ). It is possible that region IV serves as a seeding region for de novo methylation in prostate cancer cells, and that in some cases, the methylation spreads to the upstream transcription start site region, presumably causing FYN gene silencing. It remains to be finally resolved, if moderate density methylation of region IV by itself has significant influence on FYN gene expression. Our data did not reveal a clear difference between the incidence of promoter hypermethylation in tumors with or without LOH. We note, however, that neither of the most heavily methylated tumors (PC-20 and PC-57) showed LOH at the FYN locus.
In this study, we have presented several lines of evidence that point to the SRC family tyrosin kinase gene, FYN, as a new candidate TSG in human prostate adenocarcinoma. First, FYN expression was shown to be significantly downregulated at the RNA and protein level in clinical prostate cancer samples as well as in prostate adenocarcinoma cell lines compared with their nonmalignant counterparts. Second, FYN was found to be a common target for both chromosomal deletion (10/40 tumors = 25%) and dense promoter hypermethylation (2/18 tumors = 11%) in clinical prostate cancer. The identification of two independent mechanisms for downregulation/silencing of FYN in prostate cancer strongly suggests that loss of FYN provides prostate cells with a growth and/or survival advantage, and therefore may be selected during prostate cancer development.
Initially, to identify a possible candidate TSG at the frequently deleted long arm of chromosome 6 in prostate cancer, we used a novel approach that aimed to exploit the power of retroviral tagging for cancer gene discovery.9 The approach was developed upon the assumption that different tumor types can share a common tumor suppressor and that such genes may be located at regions of recurrent chromosomal loss shared by these tumors. There is substantial evidence for the existence of such shared TSGs in the literature (e.g. PTEN34). The overlap between genes disrupted by retroviral insertional mutagenesis in murine tumors and genes located at the commonly deleted region 6q14-22 in human prostate cancer, however, was fairly small (12/201 = 6%). A possible explanation is that, although several TSGs have been tagged by retroviral insertions, common integration sites are generally biased toward proto-oncogenes.9, 10 Nevertheless, FYN was successfully selected as a candidate for further analysis by this approach.
Using immunohistochemical and Western blotting analyses, we found that FYN protein expression is significantly downregulated in primary prostate cancer compared with nonmalignant prostate tissue and BPH (Figs. 3b and 3c, Fig. 4). These findings extrapolate to the protein level, the relatively low FYN mRNA levels measured in prostate cancer in several microarray-based studies.23, 24, 25, 26, 27, 28 We also measured decreased FYN expression in 6 prostate adenocarcinoma cell lines relative to nonmalignant BPH-1 cells (Figs. 2c and 2d). Moreover, our work demonstrated for the first time that prostate cells express 2 different transcript variants of FYN (FYN(B) and FYN(T)), and that both are downregulated in prostate adenocarcinoma cells (Fig. 2c).
Although the same FYN antibodies were used for Western analysis of crude extracts from cell lines and tissue samples, slightly different band patterns were observed (Figs. 2d and 3b). We are convinced that the intense smaller sized band (∼53–54 kD) in all tissue samples as well as the sporadic weaker band of ∼56 kDa (Fig. 3b) represent proteolyzed products of FYN, as previously seen by other authors.12, 29 Occasionally, we detected such additional smaller sized bands also in extracts from prostate cell lines (Fig. 2d). Hence, the different band patterns are most probably due to poor preservation of full-length FYN in extracts prepared from frozen prostate tissue samples. N-Terminal cleavage of FYN by caspase-3 and related proteases during apoptosis produces a shorter FYN protein of ∼57 kDa that relocates to the cytoplasm.14, 15 We did not, however, detect cleavage of procaspase-3 or the caspase substrate PARP by Western analysis in any of the prostate tissue samples tested (not shown). Furthermore, FYN staining was seen predominantly at the plasma membrane by immunohistochemical analysis of prostate tissue sections (Figs. 4a and 4c).
By SNP microarray analysis, LOH at the FYN locus was detected in 10 of 40 prostate cancer tissue samples (25%; Fig. 1a). Three adenocarcinoma cell lines with low FYN expression (22rv1, DU145 and PC-3) have earlier been reported to contain partial 6q deletions, affecting the FYN gene at 6q21,3, 35, 36 clearly showing that chromosomal deletion of FYN is a recurrent event in prostate cancer. The FYN gene is located at 6q21, i.e., distal to the minimal common region of LOH that centers around 6q15 (Fig. 1a), suggesting that it may not be the primary target of chromosomal deletion at 6q in prostate cancer. In a study using microsatellite markers to define in detail the deleted region at 6q in prostate cancer, however, 2 distinct regions of common loss were identified, the more distal of which mapped to 6q21-6q22.3 This suggests that another target TSG is located within this region, which contains the FYN gene.
By genomic bisulfite sequencing, we detected moderate density aberrant methylation (7–13%) of the FYN promoter CpG island in 3 (LNCaP, DU145 and PC3) of 5 prostate adenocarcinoma cell lines with reduced FYN expression, whereas the CpG island was unmethylated (<2%) in nonmalignant BPH-1 cells and in PSK-1 prostate small-cell carcinoma cells characterized by high endogenous FYN levels (Fig. 5a). Aberrant methylation was found almost exclusively in the most 3′ part of the FYN promoter associated CpG island (termed region IV). Similar methylation patterns were discovered in clinical prostate cancer samples. Generally, these patterns varied between different tumors and between individual DNA clones from the same tumor (Figs. 6 and 7). Such heterogeneity has previously been reported for epigenetically downregulated TSGs, such as CDH1 in breast cancer37 and CDKN2B in leukemia.33
Dense hypermethylation around the transcription start site (start of exon 1), as detected for FYN in 2 prostate cancer samples (PC-20 and PC-57), is known to be closely associated with gene silencing.31 In contrast, it remains uncertain if moderate density methylation of the downstream located region IV by itself has significant influence on FYN expression. Treatment of prostate adenocarcinoma cell lines with the DNA methylation inhibitor 5-aza-dC did not provide a clear answer to this question, since the drug induced FYN expression not only in DU145 and PC3 cells, with methylation of region IV, but also in VCaP cells without methylation (Fig. 5b). Indirect induction of FYN expresssion by 5-aza-dC treatment, however, could be a cell line specific effect limited to VCaP cells. Hence, the possibility remains that even moderate level methylation of region IV may influence FYN expression levels to some degree. Indeed, 6–8% methylation has been found sufficient to cause significant downregulation of the well-known TSGs CDKN2B and RB, whereas moderately higher densities (>20–40%) were associated with total silencing.32, 33 Methylation of certain CpGs in region IV could also influence FYN expression by interfering with specific transcription factor binding.31 We identified 2 subregions particularly prone to methylation in both DU145 and PC3 cells (CpG no. 11–19 and CpG no. 32–37 in region IV; Fig. 5a). Interestingly, a search for transcription factor binding motifs using MatInspector software (http://www.genomatix.de) revealed a direct overlap between the first of these subregions and 2 binding motifs for nuclear respiratory factor-1 (NRF-1), the transcriptional activation activity of which is compromised by site-specific methylation of its DNA binding site.38 The second subregion coincided with CpG-containing core consensus binding motifs for 4 different transcription factors (FLI1, ZNF143, KLF6 and HES1).
Both tumors with dense FYN promoter hypermethylation (PC-20 and PC-57) showed LOH at 6q14-16, but the deletions did not affect the FYN gene at 6q21 (Fig. 1a). An interesting possibility is that high-density promoter methylation could lead to downregulation of FYN below a critical threshold, or even to complete silencing, and thereby reduce the selection pressure for LOH at the FYN gene locus at 6q21. On the contrary, low density or absence of FYN promoter methylation could result in an increased selection pressure for LOH at 6q21 to reduce FYN expression below such a threshold. The detection of low level FYN expression in all prostate adenocarcinoma cell lines may suggest that complete silencing is not required to provide a growth and/or survival advantage to these cells. Presence of dense FYN hypermethylation in 2 of 18 clinical prostate cancer samples may be explained by selection in vivo for the loss of a certain FYN regulated process important for carcinogenesis in vivo, but not reflected in cell culture.
The consistent loss of detectable FYN protein in nonmalignant cytokeratin positive glands immediately adjacent to malignant glands (Fig. 4) may indicate that FYN promoter hypermethylation is an early event in prostate cancer development. In contrast, LOH at 6q14-22 was found predominantly in metastatic tumors in both the present (Fig. 1a) and previous studies,3, 39 suggesting that the target TSG(s) at this region regulate disease progression rather than initiation. Taking into account the multitude of processes reported to be regulated through FYN in various cell types,11, 40 it is conceivable that FYN can affect both early and late events in prostate carcinogenesis. The deleted region at 6q, however, may also contain another and yet unidentified target gene of particular importance for prostate cancer progression.
It is unknown how FYN exerts its possible tumor suppressor function in prostate cancer. Both activation and inhibition of FYN signaling have been linked to oncogenesis. For example, increased FYN activity causes dedifferentiation of melanocytes41 and is associated with increased motility of metastatic melanoma cells.42 In contrast, overexpression of FYN(B) (fused to GFP) has been shown to induce differentiation and cell cycle arrest in cultured neuroblastoma cells, consistent with the reported loss of FYN expression associated with neuroblastoma progression.16 The fact that FYN has been identified as an oncogene in some cases and a possible tumor suppressor in others may reflect a delicate balance in normal cells between growth promoting and growth inhibitory functions of FYN. Deregulated FYN expression in prostate cells may alter their response to e.g., integrins, external growth factors, death factors or hormones, all of which can affect intracellular signal transduction pathways controlled by FYN. Thus, although most prostate cancer cells express both FAS and FASL, they are generally resistant to FAS-induced apoptosis,43 a process that in T-cells is known to involve FYN.15 Finally, the exact regulatory roles of FYN may depend on which splice variants are expressed, how they are posttranslationally modified and which specific upstream and downstream signaling partners are involved.
In conclusion, we have shown that FYN is downregulated in prostate cancer and identified 2 possible mechanisms behind: chromosomal deletion and promoter hypermethylation. We note that other mechanisms, e.g., the composition of active transcription factors, most probably contribute to FYN regulation in prostate cells as well. The present report is the first to describe hypermethylation of the FYN promoter in prostate adenocarcinoma, whereas FYN has previously been reported to be epigenetically downregulated in breast17 and gastric cancer cell lines.18 Moreover, our results indicate that region IV of the FYN promoter CpG island is a preferred seeding region for de novo methylation in prostate cancer cells with epigenic control abnormalities.
The authors thank Dr. Kenneth Pienta, University of Michigan, for kindly providing the VCaP and DuCaP cells, Dr. Adrie van Boekhoven, University of Colorado, for the kind gift of PSK-1 and 1013L cells and Dr. Marilyn Resh, Memorial Sloan-Kettering Cancer Center, New York, for kindly providing the pCMV5-huFYN expression vector. The excellent technical assistance of Ms. Karen Bihl, Ms. Pamela Celis, Ms. Susanne Bruun and Ms. Lisbet Kjeldsen is gratefully acknowledged.